Based On The Theme By The Chinese Single-document Summarization System

Automatic abstracting is an important application in natural language processing area, which is also a difficult and challenging job. It has been widely used in information retrieval, information management, and digital library fields. So it will be of great theoretic value and practical significance for the research of automatic abstracting.Automatic abstracting based on statistics is an early researched and widely used method. The most advantage of this method is non-restrict of fields, and texts of different areas can use this method to get abstract. But the method has the shortcomings of incomprehensiveness, not conciseness and incoherence, which made the result of abstract can not perfectly.Based on the statistics automatic abstracting, this paper makes the two technologies of topic partition and abstract optimization into statistics automatic abstracting, which made the abstract more comprehensiveness, conciseness and coherence. The specific research contents as follows:1. Improved k-means algorithm is raised to divide the topics of text, which made the abstract more completely.2. Optimize the abstract based on the rough abstract, which made the abstract more concise and coherent.3. Design a Chinese single-document automatic abstracting system prototype based on the two steps above.Use intrinsic method to valuate the system, which include compare the system with the ideal abstract and compare the system with statistical automatic abstracting system and word2003 automatic abstracting system, and the result shows that the system is better than the other two systems.